/[pcre]/code/trunk/pcre_dfa_exec.c
ViewVC logotype

Diff of /code/trunk/pcre_dfa_exec.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 365 by ph10, Fri Jul 11 17:06:55 2008 UTC revision 518 by ph10, Tue May 18 15:47:01 2010 UTC
# Line 3  Line 3 
3  *************************************************/  *************************************************/
4    
5  /* PCRE is a library of functions to support regular expressions whose syntax  /* PCRE is a library of functions to support regular expressions whose syntax
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language (but see
7    below for why this module is different).
8    
9                         Written by Philip Hazel                         Written by Philip Hazel
10             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
11    
12  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
13  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 44  FSM). This is NOT Perl- compatible, but Line 45  FSM). This is NOT Perl- compatible, but
45  applications. */  applications. */
46    
47    
48    /* NOTE ABOUT PERFORMANCE: A user of this function sent some code that improved
49    the performance of his patterns greatly. I could not use it as it stood, as it
50    was not thread safe, and made assumptions about pattern sizes. Also, it caused
51    test 7 to loop, and test 9 to crash with a segfault.
52    
53    The issue is the check for duplicate states, which is done by a simple linear
54    search up the state list. (Grep for "duplicate" below to find the code.) For
55    many patterns, there will never be many states active at one time, so a simple
56    linear search is fine. In patterns that have many active states, it might be a
57    bottleneck. The suggested code used an indexing scheme to remember which states
58    had previously been used for each character, and avoided the linear search when
59    it knew there was no chance of a duplicate. This was implemented when adding
60    states to the state lists.
61    
62    I wrote some thread-safe, not-limited code to try something similar at the time
63    of checking for duplicates (instead of when adding states), using index vectors
64    on the stack. It did give a 13% improvement with one specially constructed
65    pattern for certain subject strings, but on other strings and on many of the
66    simpler patterns in the test suite it did worse. The major problem, I think,
67    was the extra time to initialize the index. This had to be done for each call
68    of internal_dfa_exec(). (The supplied patch used a static vector, initialized
69    only once - I suspect this was the cause of the problems with the tests.)
70    
71    Overall, I concluded that the gains in some cases did not outweigh the losses
72    in others, so I abandoned this code. */
73    
74    
75    
76  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
77  #include "config.h"  #include "config.h"
78  #endif  #endif
# Line 60  applications. */ Line 89  applications. */
89  #define SP "                   "  #define SP "                   "
90    
91    
   
92  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
94  *************************************************/  *************************************************/
# Line 78  never stored, so we push them well clear Line 106  never stored, so we push them well clear
106    
107    
108  /* This table identifies those opcodes that are followed immediately by a  /* This table identifies those opcodes that are followed immediately by a
109  character that is to be tested in some way. This makes is possible to  character that is to be tested in some way. This makes it possible to
110  centralize the loading of these characters. In the case of Type * etc, the  centralize the loading of these characters. In the case of Type * etc, the
111  "character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a  "character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a
112  small value. ***NOTE*** If the start of this table is modified, the two tables  small value. Non-zero values in the table are the offsets from the opcode where
113  that follow must also be modified. */  the character is to be found. ***NOTE*** If the start of this table is
114    modified, the three tables that follow must also be modified. */
115    
116  static const uschar coptable[] = {  static const uschar coptable[] = {
117    0,                             /* End                                    */    0,                             /* End                                    */
118    0, 0, 0, 0, 0,                 /* \A, \G, \K, \B, \b                     */    0, 0, 0, 0, 0,                 /* \A, \G, \K, \B, \b                     */
119    0, 0, 0, 0, 0, 0,              /* \D, \d, \S, \s, \W, \w                 */    0, 0, 0, 0, 0, 0,              /* \D, \d, \S, \s, \W, \w                 */
120    0, 0, 0,                       /* Any, AllAny, Anybyte                   */    0, 0, 0,                       /* Any, AllAny, Anybyte                   */
121    0, 0, 0,                       /* NOTPROP, PROP, EXTUNI                  */    0, 0,                          /* \P, \p                                 */
122    0, 0, 0, 0, 0,                 /* \R, \H, \h, \V, \v                     */    0, 0, 0, 0, 0,                 /* \R, \H, \h, \V, \v                     */
123      0,                             /* \X                                     */
124    0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */    0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */
125    1,                             /* Char                                   */    1,                             /* Char                                   */
126    1,                             /* Charnc                                 */    1,                             /* Charnc                                 */
# Line 127  static const uschar coptable[] = { Line 157  static const uschar coptable[] = {
157    0,                             /* Reverse                                */    0,                             /* Reverse                                */
158    0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */    0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */
159    0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */    0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */
160    0,                             /* CREF                                   */    0, 0,                          /* CREF, NCREF                            */
161    0,                             /* RREF                                   */    0, 0,                          /* RREF, NRREF                            */
162    0,                             /* DEF                                    */    0,                             /* DEF                                    */
163    0, 0,                          /* BRAZERO, BRAMINZERO                    */    0, 0,                          /* BRAZERO, BRAMINZERO                    */
164    0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */    0, 0, 0,                       /* MARK, PRUNE, PRUNE_ARG,                */
165    0, 0, 0                        /* FAIL, ACCEPT, SKIPZERO                 */    0, 0, 0, 0,                    /* SKIP, SKIP_ARG, THEN, THEN_ARG,        */
166      0, 0, 0, 0, 0                  /* COMMIT, FAIL, ACCEPT, CLOSE, SKIPZERO  */
167    };
168    
169    /* This table identifies those opcodes that inspect a character. It is used to
170    remember the fact that a character could have been inspected when the end of
171    the subject is reached. ***NOTE*** If the start of this table is modified, the
172    two tables that follow must also be modified. */
173    
174    static const uschar poptable[] = {
175      0,                             /* End                                    */
176      0, 0, 0, 1, 1,                 /* \A, \G, \K, \B, \b                     */
177      1, 1, 1, 1, 1, 1,              /* \D, \d, \S, \s, \W, \w                 */
178      1, 1, 1,                       /* Any, AllAny, Anybyte                   */
179      1, 1,                          /* \P, \p                                 */
180      1, 1, 1, 1, 1,                 /* \R, \H, \h, \V, \v                     */
181      1,                             /* \X                                     */
182      0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */
183      1,                             /* Char                                   */
184      1,                             /* Charnc                                 */
185      1,                             /* not                                    */
186      /* Positive single-char repeats                                          */
187      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
188      1, 1, 1,                       /* upto, minupto, exact                   */
189      1, 1, 1, 1,                    /* *+, ++, ?+, upto+                      */
190      /* Negative single-char repeats - only for chars < 256                   */
191      1, 1, 1, 1, 1, 1,              /* NOT *, *?, +, +?, ?, ??                */
192      1, 1, 1,                       /* NOT upto, minupto, exact               */
193      1, 1, 1, 1,                    /* NOT *+, ++, ?+, upto+                  */
194      /* Positive type repeats                                                 */
195      1, 1, 1, 1, 1, 1,              /* Type *, *?, +, +?, ?, ??               */
196      1, 1, 1,                       /* Type upto, minupto, exact              */
197      1, 1, 1, 1,                    /* Type *+, ++, ?+, upto+                 */
198      /* Character class & ref repeats                                         */
199      1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */
200      1, 1,                          /* CRRANGE, CRMINRANGE                    */
201      1,                             /* CLASS                                  */
202      1,                             /* NCLASS                                 */
203      1,                             /* XCLASS - variable length               */
204      0,                             /* REF                                    */
205      0,                             /* RECURSE                                */
206      0,                             /* CALLOUT                                */
207      0,                             /* Alt                                    */
208      0,                             /* Ket                                    */
209      0,                             /* KetRmax                                */
210      0,                             /* KetRmin                                */
211      0,                             /* Assert                                 */
212      0,                             /* Assert not                             */
213      0,                             /* Assert behind                          */
214      0,                             /* Assert behind not                      */
215      0,                             /* Reverse                                */
216      0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */
217      0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */
218      0, 0,                          /* CREF, NCREF                            */
219      0, 0,                          /* RREF, NRREF                            */
220      0,                             /* DEF                                    */
221      0, 0,                          /* BRAZERO, BRAMINZERO                    */
222      0, 0, 0,                       /* MARK, PRUNE, PRUNE_ARG,                */
223      0, 0, 0, 0,                    /* SKIP, SKIP_ARG, THEN, THEN_ARG,        */
224      0, 0, 0, 0, 0                  /* COMMIT, FAIL, ACCEPT, CLOSE, SKIPZERO  */
225  };  };
226    
227  /* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W,  /* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W,
# Line 170  typedef struct stateblock { Line 259  typedef struct stateblock {
259  #define INTS_PER_STATEBLOCK  (sizeof(stateblock)/sizeof(int))  #define INTS_PER_STATEBLOCK  (sizeof(stateblock)/sizeof(int))
260    
261    
262  #ifdef DEBUG  #ifdef PCRE_DEBUG
263  /*************************************************  /*************************************************
264  *             Print character string             *  *             Print character string             *
265  *************************************************/  *************************************************/
# Line 390  if (*first_op == OP_REVERSE) Line 479  if (*first_op == OP_REVERSE)
479      current_subject -= gone_back;      current_subject -= gone_back;
480      }      }
481    
482      /* Save the earliest consulted character */
483    
484      if (current_subject < md->start_used_ptr)
485        md->start_used_ptr = current_subject;
486    
487    /* Now we can process the individual branches. */    /* Now we can process the individual branches. */
488    
489    end_code = this_start_code;    end_code = this_start_code;
# Line 454  for (;;) Line 548  for (;;)
548    int i, j;    int i, j;
549    int clen, dlen;    int clen, dlen;
550    unsigned int c, d;    unsigned int c, d;
551      int forced_fail = 0;
552      BOOL could_continue = FALSE;
553    
554    /* Make the new state list into the active state list and empty the    /* Make the new state list into the active state list and empty the
555    new state list. */    new state list. */
# Line 467  for (;;) Line 563  for (;;)
563    workspace[0] ^= 1;              /* Remember for the restarting feature */    workspace[0] ^= 1;              /* Remember for the restarting feature */
564    workspace[1] = active_count;    workspace[1] = active_count;
565    
566  #ifdef DEBUG  #ifdef PCRE_DEBUG
567    printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);    printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);
568    pchars((uschar *)ptr, strlen((char *)ptr), stdout);    pchars((uschar *)ptr, strlen((char *)ptr), stdout);
569    printf("\"\n");    printf("\"\n");
# Line 511  for (;;) Line 607  for (;;)
607      stateblock *current_state = active_states + i;      stateblock *current_state = active_states + i;
608      const uschar *code;      const uschar *code;
609      int state_offset = current_state->offset;      int state_offset = current_state->offset;
610      int count, codevalue;      int count, codevalue, rrc;
611    
612  #ifdef DEBUG  #ifdef PCRE_DEBUG
613      printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);      printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);
614      if (clen == 0) printf("EOL\n");      if (clen == 0) printf("EOL\n");
615        else if (c > 32 && c < 127) printf("'%c'\n", c);        else if (c > 32 && c < 127) printf("'%c'\n", c);
# Line 543  for (;;) Line 639  for (;;)
639          }          }
640        }        }
641    
642      /* Check for a duplicate state with the same count, and skip if found. */      /* Check for a duplicate state with the same count, and skip if found.
643        See the note at the head of this module about the possibility of improving
644        performance here. */
645    
646      for (j = 0; j < i; j++)      for (j = 0; j < i; j++)
647        {        {
# Line 560  for (;;) Line 658  for (;;)
658      code = start_code + state_offset;      code = start_code + state_offset;
659      codevalue = *code;      codevalue = *code;
660    
661        /* If this opcode inspects a character, but we are at the end of the
662        subject, remember the fact for use when testing for a partial match. */
663    
664        if (clen == 0 && poptable[codevalue] != 0)
665          could_continue = TRUE;
666    
667      /* If this opcode is followed by an inline character, load it. It is      /* If this opcode is followed by an inline character, load it. It is
668      tempting to test for the presence of a subject character here, but that      tempting to test for the presence of a subject character here, but that
669      is wrong, because sometimes zero repetitions of the subject are      is wrong, because sometimes zero repetitions of the subject are
# Line 606  for (;;) Line 710  for (;;)
710    
711      switch (codevalue)      switch (codevalue)
712        {        {
713    /* ========================================================================== */
714          /* These cases are never obeyed. This is a fudge that causes a compile-
715          time error if the vectors coptable or poptable, which are indexed by
716          opcode, are not the correct length. It seems to be the only way to do
717          such a check at compile time, as the sizeof() operator does not work
718          in the C preprocessor. */
719    
720          case OP_TABLE_LENGTH:
721          case OP_TABLE_LENGTH +
722            ((sizeof(coptable) == OP_TABLE_LENGTH) &&
723             (sizeof(poptable) == OP_TABLE_LENGTH)):
724          break;
725    
726  /* ========================================================================== */  /* ========================================================================== */
727        /* Reached a closing bracket. If not at the end of the pattern, carry        /* Reached a closing bracket. If not at the end of the pattern, carry
728        on with the next opcode. Otherwise, unless we have an empty string and        on with the next opcode. Otherwise, unless we have an empty string and
729        PCRE_NOTEMPTY is set, save the match data, shifting up all previous        PCRE_NOTEMPTY is set, or PCRE_NOTEMPTY_ATSTART is set and we are at the
730          start of the subject, save the match data, shifting up all previous
731        matches so we always have the longest first. */        matches so we always have the longest first. */
732    
733        case OP_KET:        case OP_KET:
# Line 624  for (;;) Line 741  for (;;)
741            ADD_ACTIVE(state_offset - GET(code, 1), 0);            ADD_ACTIVE(state_offset - GET(code, 1), 0);
742            }            }
743          }          }
744        else if (ptr > current_subject || (md->moptions & PCRE_NOTEMPTY) == 0)        else
745          {          {
746          if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;          if (ptr > current_subject ||
747            else if (match_count > 0 && ++match_count * 2 >= offsetcount)              ((md->moptions & PCRE_NOTEMPTY) == 0 &&
748              match_count = 0;                ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 ||
749          count = ((match_count == 0)? offsetcount : match_count * 2) - 2;                  current_subject > start_subject + md->start_offset)))
750          if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int));            {
751          if (offsetcount >= 2)            if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;
752            {              else if (match_count > 0 && ++match_count * 2 >= offsetcount)
753            offsets[0] = current_subject - start_subject;                match_count = 0;
754            offsets[1] = ptr - start_subject;            count = ((match_count == 0)? offsetcount : match_count * 2) - 2;
755            DPRINTF(("%.*sSet matched string = \"%.*s\"\n", rlevel*2-2, SP,            if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int));
756              offsets[1] - offsets[0], current_subject));            if (offsetcount >= 2)
757            }              {
758          if ((md->moptions & PCRE_DFA_SHORTEST) != 0)              offsets[0] = current_subject - start_subject;
759            {              offsets[1] = ptr - start_subject;
760            DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"              DPRINTF(("%.*sSet matched string = \"%.*s\"\n", rlevel*2-2, SP,
761              "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel,                offsets[1] - offsets[0], current_subject));
762              match_count, rlevel*2-2, SP));              }
763            return match_count;            if ((md->moptions & PCRE_DFA_SHORTEST) != 0)
764                {
765                DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"
766                  "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel,
767                  match_count, rlevel*2-2, SP));
768                return match_count;
769                }
770            }            }
771          }          }
772        break;        break;
# Line 757  for (;;) Line 880  for (;;)
880        if ((md->moptions & PCRE_NOTEOL) == 0)        if ((md->moptions & PCRE_NOTEOL) == 0)
881          {          {
882          if (clen == 0 ||          if (clen == 0 ||
883              (IS_NEWLINE(ptr) &&              ((md->poptions & PCRE_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr) &&
884                 ((ims & PCRE_MULTILINE) != 0 || ptr == end_subject - md->nllen)                 ((ims & PCRE_MULTILINE) != 0 || ptr == end_subject - md->nllen)
885              ))              ))
886            { ADD_ACTIVE(state_offset + 1, 0); }            { ADD_ACTIVE(state_offset + 1, 0); }
# Line 794  for (;;) Line 917  for (;;)
917          if (ptr > start_subject)          if (ptr > start_subject)
918            {            {
919            const uschar *temp = ptr - 1;            const uschar *temp = ptr - 1;
920              if (temp < md->start_used_ptr) md->start_used_ptr = temp;
921  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
922            if (utf8) BACKCHAR(temp);            if (utf8) BACKCHAR(temp);
923  #endif  #endif
924            GETCHARTEST(d, temp);            GETCHARTEST(d, temp);
925    #ifdef SUPPORT_UCP
926              if ((md->poptions & PCRE_UCP) != 0)
927                {
928                if (d == '_') left_word = TRUE; else
929                  {
930                  int cat = UCD_CATEGORY(d);
931                  left_word = (cat == ucp_L || cat == ucp_N);
932                  }
933                }
934              else
935    #endif
936            left_word = d < 256 && (ctypes[d] & ctype_word) != 0;            left_word = d < 256 && (ctypes[d] & ctype_word) != 0;
937            }            }
938          else left_word = 0;          else left_word = FALSE;
939    
940          if (clen > 0) right_word = c < 256 && (ctypes[c] & ctype_word) != 0;          if (clen > 0)
941            else right_word = 0;            {
942    #ifdef SUPPORT_UCP
943              if ((md->poptions & PCRE_UCP) != 0)
944                {
945                if (c == '_') right_word = TRUE; else
946                  {
947                  int cat = UCD_CATEGORY(c);
948                  right_word = (cat == ucp_L || cat == ucp_N);
949                  }
950                }
951              else
952    #endif
953              right_word = c < 256 && (ctypes[c] & ctype_word) != 0;
954              }
955            else right_word = FALSE;
956    
957          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))          if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))
958            { ADD_ACTIVE(state_offset + 1, 0); }            { ADD_ACTIVE(state_offset + 1, 0); }
# Line 830  for (;;) Line 979  for (;;)
979            break;            break;
980    
981            case PT_LAMP:            case PT_LAMP:
982            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
983                   prop->chartype == ucp_Lt;
984            break;            break;
985    
986            case PT_GC:            case PT_GC:
# Line 844  for (;;) Line 994  for (;;)
994            case PT_SC:            case PT_SC:
995            OK = prop->script == code[2];            OK = prop->script == code[2];
996            break;            break;
997    
998              /* These are specials for combination cases. */
999    
1000              case PT_ALNUM:
1001              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1002                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
1003              break;
1004    
1005              case PT_SPACE:    /* Perl space */
1006              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1007                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1008              break;
1009    
1010              case PT_PXSPACE:  /* POSIX space */
1011              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1012                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1013                   c == CHAR_FF || c == CHAR_CR;
1014              break;
1015    
1016              case PT_WORD:
1017              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1018                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
1019                   c == CHAR_UNDERSCORE;
1020              break;
1021    
1022            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
1023    
# Line 999  for (;;) Line 1173  for (;;)
1173            break;            break;
1174    
1175            case PT_LAMP:            case PT_LAMP:
1176            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
1177                prop->chartype == ucp_Lt;
1178            break;            break;
1179    
1180            case PT_GC:            case PT_GC:
# Line 1014  for (;;) Line 1189  for (;;)
1189            OK = prop->script == code[3];            OK = prop->script == code[3];
1190            break;            break;
1191    
1192              /* These are specials for combination cases. */
1193    
1194              case PT_ALNUM:
1195              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1196                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
1197              break;
1198    
1199              case PT_SPACE:    /* Perl space */
1200              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1201                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1202              break;
1203    
1204              case PT_PXSPACE:  /* POSIX space */
1205              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1206                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1207                   c == CHAR_FF || c == CHAR_CR;
1208              break;
1209    
1210              case PT_WORD:
1211              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1212                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
1213                   c == CHAR_UNDERSCORE;
1214              break;
1215    
1216            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
1217    
1218            default:            default:
# Line 1221  for (;;) Line 1420  for (;;)
1420            break;            break;
1421    
1422            case PT_LAMP:            case PT_LAMP:
1423            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
1424                prop->chartype == ucp_Lt;
1425            break;            break;
1426    
1427            case PT_GC:            case PT_GC:
# Line 1235  for (;;) Line 1435  for (;;)
1435            case PT_SC:            case PT_SC:
1436            OK = prop->script == code[3];            OK = prop->script == code[3];
1437            break;            break;
1438    
1439              /* These are specials for combination cases. */
1440    
1441              case PT_ALNUM:
1442              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1443                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
1444              break;
1445    
1446              case PT_SPACE:    /* Perl space */
1447              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1448                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1449              break;
1450    
1451              case PT_PXSPACE:  /* POSIX space */
1452              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1453                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1454                   c == CHAR_FF || c == CHAR_CR;
1455              break;
1456    
1457              case PT_WORD:
1458              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1459                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
1460                   c == CHAR_UNDERSCORE;
1461              break;
1462    
1463            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
1464    
# Line 1468  for (;;) Line 1692  for (;;)
1692            break;            break;
1693    
1694            case PT_LAMP:            case PT_LAMP:
1695            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || prop->chartype == ucp_Lt;            OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll ||
1696                prop->chartype == ucp_Lt;
1697            break;            break;
1698    
1699            case PT_GC:            case PT_GC:
# Line 1482  for (;;) Line 1707  for (;;)
1707            case PT_SC:            case PT_SC:
1708            OK = prop->script == code[5];            OK = prop->script == code[5];
1709            break;            break;
1710    
1711              /* These are specials for combination cases. */
1712    
1713              case PT_ALNUM:
1714              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1715                   _pcre_ucp_gentype[prop->chartype] == ucp_N;
1716              break;
1717    
1718              case PT_SPACE:    /* Perl space */
1719              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1720                   c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1721              break;
1722    
1723              case PT_PXSPACE:  /* POSIX space */
1724              OK = _pcre_ucp_gentype[prop->chartype] == ucp_Z ||
1725                   c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1726                   c == CHAR_FF || c == CHAR_CR;
1727              break;
1728    
1729              case PT_WORD:
1730              OK = _pcre_ucp_gentype[prop->chartype] == ucp_L ||
1731                   _pcre_ucp_gentype[prop->chartype] == ucp_N ||
1732                   c == CHAR_UNDERSCORE;
1733              break;
1734    
1735            /* Should never occur, but keep compilers from grumbling. */            /* Should never occur, but keep compilers from grumbling. */
1736    
# Line 2157  for (;;) Line 2406  for (;;)
2406    
2407  /* ========================================================================== */  /* ========================================================================== */
2408        /* These are the opcodes for fancy brackets of various kinds. We have        /* These are the opcodes for fancy brackets of various kinds. We have
2409        to use recursion in order to handle them. The "always failing" assersion        to use recursion in order to handle them. The "always failing" assertion
2410        (?!) is optimised when compiling to OP_FAIL, so we have to support that,        (?!) is optimised to OP_FAIL when compiling, so we have to support that,
2411        though the other "backtracking verbs" are not supported. */        though the other "backtracking verbs" are not supported. */
2412    
2413        case OP_FAIL:        case OP_FAIL:
2414          forced_fail++;    /* Count FAILs for multiple states */
2415        break;        break;
2416    
2417        case OP_ASSERT:        case OP_ASSERT:
# Line 2189  for (;;) Line 2439  for (;;)
2439            rlevel,                               /* function recursion level */            rlevel,                               /* function recursion level */
2440            recursing);                           /* pass on regex recursion */            recursing);                           /* pass on regex recursion */
2441    
2442            if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2443          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))          if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))
2444              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
2445          }          }
# Line 2200  for (;;) Line 2451  for (;;)
2451          {          {
2452          int local_offsets[1000];          int local_offsets[1000];
2453          int local_workspace[1000];          int local_workspace[1000];
2454          int condcode = code[LINK_SIZE+1];          int codelink = GET(code, 1);
2455            int condcode;
2456    
2457            /* Because of the way auto-callout works during compile, a callout item
2458            is inserted between OP_COND and an assertion condition. This does not
2459            happen for the other conditions. */
2460    
2461            if (code[LINK_SIZE+1] == OP_CALLOUT)
2462              {
2463              rrc = 0;
2464              if (pcre_callout != NULL)
2465                {
2466                pcre_callout_block cb;
2467                cb.version          = 1;   /* Version 1 of the callout block */
2468                cb.callout_number   = code[LINK_SIZE+2];
2469                cb.offset_vector    = offsets;
2470                cb.subject          = (PCRE_SPTR)start_subject;
2471                cb.subject_length   = end_subject - start_subject;
2472                cb.start_match      = current_subject - start_subject;
2473                cb.current_position = ptr - start_subject;
2474                cb.pattern_position = GET(code, LINK_SIZE + 3);
2475                cb.next_item_length = GET(code, 3 + 2*LINK_SIZE);
2476                cb.capture_top      = 1;
2477                cb.capture_last     = -1;
2478                cb.callout_data     = md->callout_data;
2479                if ((rrc = (*pcre_callout)(&cb)) < 0) return rrc;   /* Abandon */
2480                }
2481              if (rrc > 0) break;                      /* Fail this thread */
2482              code += _pcre_OP_lengths[OP_CALLOUT];    /* Skip callout data */
2483              }
2484    
2485            condcode = code[LINK_SIZE+1];
2486    
2487          /* Back reference conditions are not supported */          /* Back reference conditions are not supported */
2488    
2489          if (condcode == OP_CREF) return PCRE_ERROR_DFA_UCOND;          if (condcode == OP_CREF || condcode == OP_NCREF)
2490              return PCRE_ERROR_DFA_UCOND;
2491    
2492          /* The DEFINE condition is always false */          /* The DEFINE condition is always false */
2493    
2494          if (condcode == OP_DEF)          if (condcode == OP_DEF)
2495            {            { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
           ADD_ACTIVE(state_offset + GET(code, 1) + LINK_SIZE + 1, 0);  
           }  
2496    
2497          /* The only supported version of OP_RREF is for the value RREF_ANY,          /* The only supported version of OP_RREF is for the value RREF_ANY,
2498          which means "test if in any recursion". We can't test for specifically          which means "test if in any recursion". We can't test for specifically
2499          recursed groups. */          recursed groups. */
2500    
2501          else if (condcode == OP_RREF)          else if (condcode == OP_RREF || condcode == OP_NRREF)
2502            {            {
2503            int value = GET2(code, LINK_SIZE+2);            int value = GET2(code, LINK_SIZE+2);
2504            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;            if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;
2505            if (recursing > 0) { ADD_ACTIVE(state_offset + LINK_SIZE + 4, 0); }            if (recursing > 0)
2506              else { ADD_ACTIVE(state_offset + GET(code, 1) + LINK_SIZE + 1, 0); }              { ADD_ACTIVE(state_offset + LINK_SIZE + 4, 0); }
2507              else { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
2508            }            }
2509    
2510          /* Otherwise, the condition is an assertion */          /* Otherwise, the condition is an assertion */
# Line 2248  for (;;) Line 2530  for (;;)
2530              rlevel,                               /* function recursion level */              rlevel,                               /* function recursion level */
2531              recursing);                           /* pass on regex recursion */              recursing);                           /* pass on regex recursion */
2532    
2533              if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2534            if ((rc >= 0) ==            if ((rc >= 0) ==
2535                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))                  (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))
2536              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }              { ADD_ACTIVE(endasscode + LINK_SIZE + 1 - start_code, 0); }
2537            else            else
2538              { ADD_ACTIVE(state_offset + GET(code, 1) + LINK_SIZE + 1, 0); }              { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
2539            }            }
2540          }          }
2541        break;        break;
# Line 2404  for (;;) Line 2687  for (;;)
2687        /* Handle callouts */        /* Handle callouts */
2688    
2689        case OP_CALLOUT:        case OP_CALLOUT:
2690          rrc = 0;
2691        if (pcre_callout != NULL)        if (pcre_callout != NULL)
2692          {          {
         int rrc;  
2693          pcre_callout_block cb;          pcre_callout_block cb;
2694          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 1;   /* Version 1 of the callout block */
2695          cb.callout_number   = code[1];          cb.callout_number   = code[1];
# Line 2421  for (;;) Line 2704  for (;;)
2704          cb.capture_last     = -1;          cb.capture_last     = -1;
2705          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
2706          if ((rrc = (*pcre_callout)(&cb)) < 0) return rrc;   /* Abandon */          if ((rrc = (*pcre_callout)(&cb)) < 0) return rrc;   /* Abandon */
         if (rrc == 0) { ADD_ACTIVE(state_offset + 2 + 2*LINK_SIZE, 0); }  
2707          }          }
2708          if (rrc == 0)
2709            { ADD_ACTIVE(state_offset + _pcre_OP_lengths[OP_CALLOUT], 0); }
2710        break;        break;
2711    
2712    
# Line 2438  for (;;) Line 2722  for (;;)
2722    /* We have finished the processing at the current subject character. If no    /* We have finished the processing at the current subject character. If no
2723    new states have been set for the next character, we have found all the    new states have been set for the next character, we have found all the
2724    matches that we are going to find. If we are at the top level and partial    matches that we are going to find. If we are at the top level and partial
2725    matching has been requested, check for appropriate conditions. */    matching has been requested, check for appropriate conditions.
2726    
2727      The "forced_ fail" variable counts the number of (*F) encountered for the
2728      character. If it is equal to the original active_count (saved in
2729      workspace[1]) it means that (*F) was found on every active state. In this
2730      case we don't want to give a partial match.
2731    
2732      The "could_continue" variable is true if a state could have continued but
2733      for the fact that the end of the subject was reached. */
2734    
2735    if (new_count <= 0)    if (new_count <= 0)
2736      {      {
2737      if (match_count < 0 &&                     /* No matches found */      if (rlevel == 1 &&                               /* Top level, and */
2738          rlevel == 1 &&                         /* Top level match function */          could_continue &&                            /* Some could go on */
2739          (md->moptions & PCRE_PARTIAL) != 0 &&  /* Want partial matching */          forced_fail != workspace[1] &&               /* Not all forced fail & */
2740          ptr >= end_subject &&                  /* Reached end of subject */          (                                            /* either... */
2741          ptr > current_subject)                 /* Matched non-empty string */          (md->moptions & PCRE_PARTIAL_HARD) != 0      /* Hard partial */
2742            ||                                           /* or... */
2743            ((md->moptions & PCRE_PARTIAL_SOFT) != 0 &&  /* Soft partial and */
2744             match_count < 0)                            /* no matches */
2745            ) &&                                         /* And... */
2746            ptr >= end_subject &&                     /* Reached end of subject */
2747            ptr > current_subject)                    /* Matched non-empty string */
2748        {        {
2749        if (offsetcount >= 2)        if (offsetcount >= 2)
2750          {          {
2751          offsets[0] = current_subject - start_subject;          offsets[0] = md->start_used_ptr - start_subject;
2752          offsets[1] = end_subject - start_subject;          offsets[1] = end_subject - start_subject;
2753          }          }
2754        match_count = PCRE_ERROR_PARTIAL;        match_count = PCRE_ERROR_PARTIAL;
# Line 2592  md->start_code = (const uschar *)argumen Line 2890  md->start_code = (const uschar *)argumen
2890      re->name_table_offset + re->name_count * re->name_entry_size;      re->name_table_offset + re->name_count * re->name_entry_size;
2891  md->start_subject = (const unsigned char *)subject;  md->start_subject = (const unsigned char *)subject;
2892  md->end_subject = end_subject;  md->end_subject = end_subject;
2893    md->start_offset = start_offset;
2894  md->moptions = options;  md->moptions = options;
2895  md->poptions = re->options;  md->poptions = re->options;
2896    
# Line 2614  switch ((((options & PCRE_NEWLINE_BITS) Line 2913  switch ((((options & PCRE_NEWLINE_BITS)
2913           PCRE_NEWLINE_BITS)           PCRE_NEWLINE_BITS)
2914    {    {
2915    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Compile-time default */
2916    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
2917    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
2918    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
2919         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
2920    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
2921    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
2922    default: return PCRE_ERROR_BADNEWLINE;    default: return PCRE_ERROR_BADNEWLINE;
# Line 2696  if (!anchored) Line 2995  if (!anchored)
2995      }      }
2996    else    else
2997      {      {
2998      if (startline && study != NULL &&      if (!startline && study != NULL &&
2999           (study->options & PCRE_STUDY_MAPPED) != 0)           (study->flags & PCRE_STUDY_MAPPED) != 0)
3000        start_bits = study->start_bits;        start_bits = study->start_bits;
3001      }      }
3002    }    }
# Line 2713  if ((re->flags & PCRE_REQCHSET) != 0) Line 3012  if ((re->flags & PCRE_REQCHSET) != 0)
3012    }    }
3013    
3014  /* Call the main matching function, looping for a non-anchored regex after a  /* Call the main matching function, looping for a non-anchored regex after a
3015  failed match. Unless restarting, optimize by moving to the first match  failed match. If not restarting, perform certain optimizations at the start of
3016  character if possible, when not anchored. Then unless wanting a partial match,  a match. */
 check for a required later character. */  
3017    
3018  for (;;)  for (;;)
3019    {    {
# Line 2725  for (;;) Line 3023  for (;;)
3023      {      {
3024      const uschar *save_end_subject = end_subject;      const uschar *save_end_subject = end_subject;
3025    
3026      /* Advance to a unique first char if possible. If firstline is TRUE, the      /* If firstline is TRUE, the start of the match is constrained to the first
3027      start of the match is constrained to the first line of a multiline string.      line of a multiline string. Implement this by temporarily adjusting
3028      Implement this by temporarily adjusting end_subject so that we stop      end_subject so that we stop scanning at a newline. If the match fails at
3029      scanning at a newline. If the match fails at the newline, later code breaks      the newline, later code breaks this loop. */
     this loop. */  
3030    
3031      if (firstline)      if (firstline)
3032        {        {
3033        USPTR t = current_subject;        USPTR t = current_subject;
3034  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3035        if (utf8)        if (utf8)
3036          {          {
3037          while (t < md->end_subject && !IS_NEWLINE(t))          while (t < md->end_subject && !IS_NEWLINE(t))
3038            {            {
3039            t++;            t++;
3040            while (t < end_subject && (*t & 0xc0) == 0x80) t++;            while (t < end_subject && (*t & 0xc0) == 0x80) t++;
3041            }            }
3042          }          }
3043        else        else
3044  #endif  #endif
3045        while (t < md->end_subject && !IS_NEWLINE(t)) t++;        while (t < md->end_subject && !IS_NEWLINE(t)) t++;
3046        end_subject = t;        end_subject = t;
3047        }        }
3048    
3049      if (first_byte >= 0)      /* There are some optimizations that avoid running the match if a known
3050        starting point is not found. However, there is an option that disables
3051        these, for testing and for ensuring that all callouts do actually occur. */
3052    
3053        if ((options & PCRE_NO_START_OPTIMIZE) == 0)
3054        {        {
3055        if (first_byte_caseless)        /* Advance to a known first byte. */
         while (current_subject < end_subject &&  
                lcc[*current_subject] != first_byte)  
           current_subject++;  
       else  
         while (current_subject < end_subject && *current_subject != first_byte)  
           current_subject++;  
       }  
3056    
3057      /* Or to just after a linebreak for a multiline match if possible */        if (first_byte >= 0)
3058            {
3059            if (first_byte_caseless)
3060              while (current_subject < end_subject &&
3061                     lcc[*current_subject] != first_byte)
3062                current_subject++;
3063            else
3064              while (current_subject < end_subject &&
3065                     *current_subject != first_byte)
3066                current_subject++;
3067            }
3068    
3069      else if (startline)        /* Or to just after a linebreak for a multiline match if possible */
3070        {  
3071        if (current_subject > md->start_subject + start_offset)        else if (startline)
3072          {          {
3073  #ifdef SUPPORT_UTF8          if (current_subject > md->start_subject + start_offset)
         if (utf8)  
3074            {            {
3075            while (current_subject < end_subject && !WAS_NEWLINE(current_subject))  #ifdef SUPPORT_UTF8
3076              if (utf8)
3077              {              {
3078              current_subject++;              while (current_subject < end_subject &&
3079              while(current_subject < end_subject &&                     !WAS_NEWLINE(current_subject))
3080                    (*current_subject & 0xc0) == 0x80)                {
3081                current_subject++;                current_subject++;
3082              }                while(current_subject < end_subject &&
3083                        (*current_subject & 0xc0) == 0x80)
3084                    current_subject++;
3085                  }
3086                }
3087              else
3088    #endif
3089              while (current_subject < end_subject && !WAS_NEWLINE(current_subject))
3090                current_subject++;
3091    
3092              /* If we have just passed a CR and the newline option is ANY or
3093              ANYCRLF, and we are now at a LF, advance the match position by one
3094              more character. */
3095    
3096              if (current_subject[-1] == CHAR_CR &&
3097                   (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&
3098                   current_subject < end_subject &&
3099                   *current_subject == CHAR_NL)
3100                current_subject++;
3101            }            }
         else  
 #endif  
         while (current_subject < end_subject && !WAS_NEWLINE(current_subject))  
           current_subject++;  
   
         /* If we have just passed a CR and the newline option is ANY or  
         ANYCRLF, and we are now at a LF, advance the match position by one more  
         character. */  
   
         if (current_subject[-1] == '\r' &&  
              (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&  
              current_subject < end_subject &&  
              *current_subject == '\n')  
           current_subject++;  
3102          }          }
       }  
3103    
3104      /* Or to a non-unique first char after study */        /* Or to a non-unique first char after study */
3105    
3106      else if (start_bits != NULL)        else if (start_bits != NULL)
       {  
       while (current_subject < end_subject)  
3107          {          {
3108          register unsigned int c = *current_subject;          while (current_subject < end_subject)
3109          if ((start_bits[c/8] & (1 << (c&7))) == 0) current_subject++;            {
3110            else break;            register unsigned int c = *current_subject;
3111              if ((start_bits[c/8] & (1 << (c&7))) == 0) current_subject++;
3112                else break;
3113              }
3114          }          }
3115        }        }
3116    
3117      /* Restore fudged end_subject */      /* Restore fudged end_subject */
3118    
3119      end_subject = save_end_subject;      end_subject = save_end_subject;
     }  
   
   /* If req_byte is set, we know that that character must appear in the subject  
   for the match to succeed. If the first character is set, req_byte must be  
   later in the subject; otherwise the test starts at the match point. This  
   optimization can save a huge amount of work in patterns with nested unlimited  
   repeats that aren't going to match. Writing separate code for cased/caseless  
   versions makes it go faster, as does using an autoincrement and backing off  
   on a match.  
   
   HOWEVER: when the subject string is very, very long, searching to its end can  
   take a long time, and give bad performance on quite ordinary patterns. This  
   showed up when somebody was matching /^C/ on a 32-megabyte string... so we  
   don't do this when the string is sufficiently long.  
   
   ALSO: this processing is disabled when partial matching is requested.  
   */  
   
   if (req_byte >= 0 &&  
       end_subject - current_subject < REQ_BYTE_MAX &&  
       (options & PCRE_PARTIAL) == 0)  
     {  
     register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);  
3120    
3121      /* We don't need to repeat the search if we haven't yet reached the      /* The following two optimizations are disabled for partial matching or if
3122      place we found it at last time. */      disabling is explicitly requested (and of course, by the test above, this
3123        code is not obeyed when restarting after a partial match). */
3124    
3125      if (p > req_byte_ptr)      if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&
3126            (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0)
3127        {        {
3128        if (req_byte_caseless)        /* If the pattern was studied, a minimum subject length may be set. This
3129          {        is a lower bound; no actual string of that length may actually match the
3130          while (p < end_subject)        pattern. Although the value is, strictly, in characters, we treat it as
3131            {        bytes to avoid spending too much time in this optimization. */
3132            register int pp = *p++;  
3133            if (pp == req_byte || pp == req_byte2) { p--; break; }        if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
3134            }            (pcre_uint32)(end_subject - current_subject) < study->minlength)
3135          }          return PCRE_ERROR_NOMATCH;
3136        else  
3137          /* If req_byte is set, we know that that character must appear in the
3138          subject for the match to succeed. If the first character is set, req_byte
3139          must be later in the subject; otherwise the test starts at the match
3140          point. This optimization can save a huge amount of work in patterns with
3141          nested unlimited repeats that aren't going to match. Writing separate
3142          code for cased/caseless versions makes it go faster, as does using an
3143          autoincrement and backing off on a match.
3144    
3145          HOWEVER: when the subject string is very, very long, searching to its end
3146          can take a long time, and give bad performance on quite ordinary
3147          patterns. This showed up when somebody was matching /^C/ on a 32-megabyte
3148          string... so we don't do this when the string is sufficiently long. */
3149    
3150          if (req_byte >= 0 && end_subject - current_subject < REQ_BYTE_MAX)
3151          {          {
3152          while (p < end_subject)          register const uschar *p = current_subject + ((first_byte >= 0)? 1 : 0);
3153    
3154            /* We don't need to repeat the search if we haven't yet reached the
3155            place we found it at last time. */
3156    
3157            if (p > req_byte_ptr)
3158            {            {
3159            if (*p++ == req_byte) { p--; break; }            if (req_byte_caseless)
3160            }              {
3161          }              while (p < end_subject)
3162                  {
3163                  register int pp = *p++;
3164                  if (pp == req_byte || pp == req_byte2) { p--; break; }
3165                  }
3166                }
3167              else
3168                {
3169                while (p < end_subject)
3170                  {
3171                  if (*p++ == req_byte) { p--; break; }
3172                  }
3173                }
3174    
3175        /* If we can't find the required character, break the matching loop,            /* If we can't find the required character, break the matching loop,
3176        which will cause a return or PCRE_ERROR_NOMATCH. */            which will cause a return or PCRE_ERROR_NOMATCH. */
3177    
3178        if (p >= end_subject) break;            if (p >= end_subject) break;
3179    
3180        /* If we have found the required character, save the point where we            /* If we have found the required character, save the point where we
3181        found it, so that we don't search again next time round the loop if            found it, so that we don't search again next time round the loop if
3182        the start hasn't passed this character yet. */            the start hasn't passed this character yet. */
3183    
3184        req_byte_ptr = p;            req_byte_ptr = p;
3185              }
3186            }
3187        }        }
3188      }      }   /* End of optimizations that are done when not restarting */
3189    
3190    /* OK, now we can do the business */    /* OK, now we can do the business */
3191    
3192      md->start_used_ptr = current_subject;
3193    
3194    rc = internal_dfa_exec(    rc = internal_dfa_exec(
3195      md,                                /* fixed match data */      md,                                /* fixed match data */
3196      md->start_code,                    /* this subexpression's code */      md->start_code,                    /* this subexpression's code */
# Line 2903  for (;;) Line 3225  for (;;)
3225    not contain any explicit matches for \r or \n, and the newline option is CRLF    not contain any explicit matches for \r or \n, and the newline option is CRLF
3226    or ANY or ANYCRLF, advance the match position by one more character. */    or ANY or ANYCRLF, advance the match position by one more character. */
3227    
3228    if (current_subject[-1] == '\r' &&    if (current_subject[-1] == CHAR_CR &&
3229        current_subject < end_subject &&        current_subject < end_subject &&
3230        *current_subject == '\n' &&        *current_subject == CHAR_NL &&
3231        (re->flags & PCRE_HASCRORLF) == 0 &&        (re->flags & PCRE_HASCRORLF) == 0 &&
3232          (md->nltype == NLTYPE_ANY ||          (md->nltype == NLTYPE_ANY ||
3233           md->nltype == NLTYPE_ANYCRLF ||           md->nltype == NLTYPE_ANYCRLF ||

Legend:
Removed from v.365  
changed lines
  Added in v.518

  ViewVC Help
Powered by ViewVC 1.1.5